October 1999
Rocky Mountain Research Station
Station
Improving Future Fluvial Classification Systems
by Craig N. Goodwin
Why do people classify things?
Psychologists theorize that in a universe of
limitless numbers of objects and ideas,
classifying things into groups is one of the
brain’s mechanisms for creating order out
of chaos. Historically, most fields of science
have gone through a classification phase.
The classification phase usually occurs
during the early stages of development of a
scientific field as a means of ordering
observations and descriptions. As a science
advances, classification gives way to the
development of empirically based laws and
finally to theoretical understanding
(Figure 1).
relatively young field that is founded upon
field studies and observation. Eventually,
however, “classifications defined by
reference to manifest, observable
characteristics will tend to give way to
systems based on theoretical concepts”
(Hempel, 1965). In fact, because scientific
usefulness is measured in part by an ability
to predict, a “preoccupation with description
could lead to decreasing usefulness because
classification and description are usually
insufficient bases for extrapolation and thus
for prediction” (Leopold and Langbein,
1963).
A Few Fluvial Classification Systems
Classification has and will continue to have
an important role in science, particularly in
fluvial geomorphology because it is a
Observation
C lassification
Empirical laws
Theory
Figure 1. Evolutionary stages in the
development of a field of science.
Form-based or morphological classification
schemes have been popular for river
classification. Many of these have used the
global property of river shape in plan form
as the primary delimiter. The best known
river classification of this type is the
tripartite division of rivers into braided,
meandering, and straight (Leopold and
Wolman, 1957).
Rosgen’s (1994)
classification of natural rivers also places a
heavy emphasis upon dimensional
properties to define eight primary stream
types. A hierarchical decision tree
distinguishes types based upon the feature
property of number of channel threads and
the dimensional properties of entrenchment
ratio, width-depth ratio, and sinuosity.
STREAM NOTES is produced
quarterly by the Stream Systems
Technology Center, Rocky
Mountain Research Station,
Fort Collins, Colorado.
The PRIMARY AIM is to exchange
technical ideas and transfer
technology among scientists
working with wildland stream
systems.
CONTRIBUTIONS are voluntary
and will be accepted at any time.
They should be typewritten, singlespaced, and limited to two pages.
Graphics and tables are
encouraged.
Ideas and opinions expressed are
not necessarily Forest Service
policy. Citations, reviews, and use
of trade names does not constitute
endorsement by the USDA Forest
Service.
Phone: (970) 498-1731
FAX: (970) 498-1660
E-Mail:stream/rmrs@ fs.fed.us
Web Site: www.stream.fs.fed.us
IN THIS ISSUE
• Improving Future
Fluvial
Classification
Systems
• Why Do We
Exaggerate
Stream Channel
Cross-Section
Plots? The Case
for True Scale
Plotting
Entrenchment ratio, sinuosity, and width-depth
ratio have fuzzy decision rules that acknowledge
the continuum of stream variability. Sediment size
and channel slope are used to classify these types
into subcategories.
An example of a process-based classification
system is that proposed by Whiting and Bradley
(1993) which utilizes dimensional measures of
fluvial form in their classification system for
headwater streams. Forty-two stream classes are
defined on the basis of domains in three twodimensional phase space “panels” where the
domains represent different and distinct physical
processes and their relative rates. The domains
are process interpretations based upon dimensional
properties of morphological features, which include
channel gradient, channel width, valley width, and
median sediment size.
2. Base Classifications upon Processes or
Controlling Factors
Many fluvial classification schemes have been based
wholly or in part upon characteristics of channel
form. Channel forms, although readily measurable,
are the end products of complex, dynamic systems
(Figure 2). These end products may be non-unique
manifestations of underlying controlling factors and
processes. The lack of a one-to-one correspondence
between geomorphic process and form suggests that
measurement of fluvial processes or controlling
factors, albeit difficult, may be a better pathway to
discovering natural kinds of rivers.
Controlling
Inputs
W a te r d is c h a rg e
S e d im e n t d is c h a rg e
Constra ints
Montgomery and Buffington’s (1997) classification
system defines seven channel types based upon
overall qualitative morphological character. Seven
fuzzy, mainly feature characteristics are used to
define whether or not a stream is of a given type.
Montgomery and Buffington’s use of adjective
feature descriptors of typical and dominant
illustrate that these are desirable features but not
absolutely necessary for describing a type. The
system uses illustration to communicate the abstract
conceptualizations of their types
Improving Fluvial Classification Systems
Within this section I offer ten recommendations to
improve future fluvial classification systems.
1. Base Classifications on Natural Kinds
Assuming there are natural kinds (Collier, 1996)
of rivers, these kinds (natural groupings as opposed
to human abstractions) should be intimately linked
with natural laws of river development. Thus,
classification systems based on natural kinds should
have utility in explaining and predicting river
conditions. The goal for the classifier is to
determine what those natural kinds are.
Va lle y s lo p e & w id th
B o u n d a ry co n d itio n s
Fluvial
Erosion
S e dim e nt Tra nsport
De position
Fluvial
Cha nne l type
Re a ch fe a ture s
Figure 2. Simple conceptualization of relationship
among controlling variables, fluvial processes, and
fluvial forms. Most classification systems have
concentrated on fluvial forms, which are the output
products of a complex, dynamic system.
3. Base Classifications upon Temporal
Change and Thresholds
Extend the process-based idea presented in
recommendation 2 to longer temporal spans wherein
rivers are truly perceived not as “things in space”
but as “processes through time.” These types of
classification systems might include concepts of
equilibrium and disequilibrium showing departures
from a steady state and could include measures such
as the ratio of reaction time to relaxation time or
threshold parameters.
4. Base Classifications on Theory
Many fields of science begin with an observationbased classification stage and then advance to
law-forming and theory development stages.
After some period of growth in a scientific field,
all of these stages occur simultaneously, with
feedbacks from more advanced levels of
knowledge guiding less advanced levels (Figure
1). Thus, at some evolutionary point, even
observations are based upon theory and not just
upon mindless collection of data. With time,
classification should become less dependent upon
purely observational input and more dependent
upon geomorphological theory.
5. Base Classifications on a
Probabilistic View
Most classification systems in geomorphology
and other scientific fields have been based upon
the classical view in which members are placed
into categories based upon “necessary and
sufficient” conditions. A more recently
developed probabilistic theory does not require
all properties to be true of all class members.
Class members vary in the degree both to which
they share a property and to which they represent
a class. Probabilistic classification concepts
appear to hold much promise, particularly since
necessary and sufficient conditions of the
classical view are replaced by central tendency
conditions, which provide more flexibility for the
complex continuum of the fluvial system.
Although none of the existing fluvial
classification systems implement a probabilistic
view, Montgomery and Buffington’s (1997)
system is a step in that direction.
6. Calibrate and Verify Classifications
for Prediction
Recommendations 1, 2, 3, and 4 intertwine
classification with scientific understanding and
explanation of fluvial conditions. If the goal of
using a classification system is to make
management predictions (as opposed to
increasing understanding), then a purely
empirical rather than rationale approach to
classification can be taken. If empirical
predictive classification schemes are to be developed
and used, then they should be calibrated, verified,
and updated as more outcome data become available.
7. Incorporate Size Factors into
Classification Systems
Nearly all existing fluvial classification schemes
ignore size and scale aspects in their derivation.
Although size measurements may be made to derive
classification parameters, these measurements are
usually converted into dimensionless shape or
pattern indicators (e.g., width-depth ratio and
sinuosity). Scale issues, however, are significant
both in absolute and in relative ways, and
classification systems that effectively incorporate
size factors may prove beneficial in endeavors such
as stream restoration.
8. Use Nomenclature that Improves
Communication
The fluvial environment is being examined by
professionals from many disciplines. For
individuals from diverse interests to effectively
describe and communicate the aspects of a complex
natural system, a classification system that “creates
order out of chaos” is beneficial. Classifications
devised to aid communication need not be based
upon natural kinds, for the purpose is merely to allow
individuals to develop mental pictures of rivers. For
this reason, form-based classification systems are
most useful, for forms are more intuitive and
understandable to humans than are geomorphic
processes. A key aspect of a classification scheme
of this type should be its composition of a readily
understandable nomenclature. The following
nomenclature rules, adapted from Hill (1963), are
suggested:
·
·
·
·
·
A name should be widely understood and used.
A name should be descriptive or explanatory.
A name should be a common word, if possible.
A name should be rational and appropriate to
the science involved.
Symbolic and/or mnemonic terms may be used
if they are more practical than descriptive terms.
· The same thing should not be given two different
names nor different things given the same name.
· A name should represent a group of things,
processes, or concepts, and if possible, should
also be part of a greater group.
classification systems. However, classification
should be considered only one part of a much larger
scientific puzzle that also incorporates observation,
laws, hypotheses, theories, and models.
References
9. Treat Classifications as Hypotheses —
Not as Paradigms
In some respects, classification schemes can be
treated similarly to hypotheses in that their
development is undertaken to seek to explain
regularities in nature. If classification systems are
viewed as hypotheses with respect to their use, then
each additional use becomes a test either verifying
or nullifying their explanatory capability. As
evidence amasses regarding a classification
scheme’s explanatory or predictive capability,
modifications can be made, and, if necessary, it can
be discarded in favor of a new system.
Unfortunately, theories, and perhaps classification
schemes, can become paradigms that are the only
way a field of science evaluates a particular class of
problems. Classification paradigms should be
avoided for a science to continue advancing.
10. Ignore Classification
Much of the desire to classify rivers may derive from
fluvial geomorphology’s composite heritage from
geology and geography — fields where observation,
description, and classification have played major
roles. In other disciplines, classification plays little
or no part. As geomorphological theory advances,
conceptual and mathematical geomorphological
models will undoubtedly provide capabilities not
inherent in classification systems.
Conclusion
Classification has and will continue to have an
important role in science, particularly for fluvial
geomorphology, which is solidly founded upon field
studies and observation. A fuller understanding of
the principles behind classification system
development should improve the design of future
Collier, J., 1996. On the Necessity of Natural Kinds. In:
Riggs, P. J. (ed.), Natural Kinds, Laws of Nature and
Scientific Methodology. Kluwer Academic
Publishers, Dordrecht, pp. 1-10.
Hempel, C. G., 1965. Aspects of Scientific Explanation.
The Free Press, New York, 505 pp.
Hill, M. L., 1963. Role of Classification in Geology. In:
Albritton, C. C. (ed.), The Fabric of Geology.
Addison-Wesley, Reading, Mass., pp. 164-174.
Leopold, L. B. and W. B. Langbein, 1963. Association
and Indeterminacy in Geomorphology. In: Albritton,
C. C. (ed.), The Fabric of Geology. Addison-Wesley,
Reading, Mass., pp. 184-192.
Leopold, L. B. and M. G. Wolman, 1957. River Channel
Patterns: Braided, Meandering and Straight. U.S.
Geological Survey Professional Paper 282-B, pp. 39­
85.
Montgomery, D. R. and J. M. Buffington, 1997. ChannelReach Morphology in Mountain Drainage Basins.
Geological Society of America Bulletin 109:596-611.
Rosgen, D. L., 1994. A Classification of Natural Rivers.
Catena 22:169-199.
Whiting, P. J. and J. B. Bradley, 1993. A Process-Based
Classification System for Headwater Streams. Earth
Surface Processes and Landforms 18:603-612.
Craig Goodwin is a geomorphologist and a
Doctoral Student/Research Associate,
Watershed Science Unit/Rangeland Resources,
Utah State University, Logan, Utah 84322­
5230; (435) 797-1587; goodwin@cc.usu.edu.
The complete article, Goodwin, Craig N., 1999.
Fluvial Classification: Neanderthal Necessity or
Needless Normalcy. In: D.S. Olson and J.P.
Potyondy (editors), Wildland Hydrology,
American Water Resources Association, TPS­
99-3, Herndon, Virginia, pp. 229-236, may be
dowloaded from the STREAM Web site
(www.stream.fs.fed.us) FTP download area.
Why Do We Exaggerate Stream Channel Cross-Section Plots?
The Case for True Scale Plotting
by John Potyondy and Larry Schmidt
Perception often governs reality. The way people
perceive objects has a profound influence on the way
they conceptualize the objects. This is true for
tangible objects that we identify with our senses, but
it applies equally well to concepts and ideas. This
article illustrates how typical displays of stream
channel cross-section data confuse perception and
lead to misinterpretation of stream channels and how
they operate.
Most people recognize the characteristics of a square.
It is a familiar basic geometric shape learned at an
early age. But suppose for a moment that you must
illustrate a square for a person unfamiliar with the
concept. Your basic description might say, for
example, that a square is a rectangle with four equal
sides (Figure 1).
Now consider the effect on the person’s perception
of a square if you provided an illustration of the
square with a 2 to 1 vertical exaggeration (Figure
2). Even though the exaggeration is clearly
identified, it is likely that the person will carry a
warped view of what a square is.
A rational person would reject the notion that this
is a good way to illustrate a square and would
consider it a poor way to teach a person the
concept of a square. As absurd as this example
seems, this is the typical case when plotting and
displaying stream channel cross-sections.
Consider the following example of a cross-section
plot of the South Fork Cache la Poudre River in
Colorado. The South Fork Cache la Poudre is a
typical mountain gravel-bed river, about 40 feet
wide and 3 feet deep at bankfull stage. The river
flows through a broad alluvial valley meadow and
can be characterized as a Rosgen C-3 stream type.
The plot in Figure 3 illustrates how the crosssection is typically displayed when plotted with
a standard computer spreadsheet, such as Excel.
One problem with this display is that it does not
show the cross-section as it appears in the natural
world. This is because vertical elevation and
horizontal distance are not plotted at identical
scales. For the cross-section to appear as it does
in reality, it would have to be plotted at a 1:1
vertical-to-horizontal ratio. In this example, the
computer graphing software exaggerated the
vertical scale by a factor of 5 giving it a 5:1
vertical-to-horizontal ratio. In other words, each
foot on the vertical scale is 5 times that of the
distance used to show 1 foot on the horizontal
scale.
In contrast, Figure 4 shows a true representation
of the actual stream cross-section (plotted at a 1:1
ratio); one that is close to how we actually see
and perceive the stream were we to stand on its
banks.
Figure 1. A square.
Figure 2. A square with
2:1 vertical
exaggeration.
While the astute viewer may be aware of the
exaggerated nature of the plot in Figure 3, the
casual viewer will probably draw different
conclusions about river behavior from the
exaggerated versus the real-life scale plot.
Figure 3. Plot of the South Fork Cache la Poudre River cross-section as the data is directly plotted by a
typical spreadsheet plotting program. The vertical axis is exaggerated by a factor of 5 making for a 5:1
vertical-to-horizontal plotting ratio. The exaggerated nature of the plot conveys a distorted view of the
river’s true cross-sectional characteristics.
Figure 4. Plot of South Fork Cache la Poudre River cross-section at a 1:1 ratio without vertical
exaggeration.
Figure 5. Panoramic view of the South Fork of the Cache la Poudre River cross-section.
In the exaggerated plot (Figure 3), the South Fork
Cache la Poudre River appears to be a relatively deep
river. Depth appears to change dramatically over
short distances across the channel and one might be
tempted to conceptually think of this river as one
which is fast and deep. Because of the large vertical
distances between low base flows and flows which
fill the channel, it looks like the river fluctuates a
great deal between wet and dry seasons. The
elevation of bankfull stage is identified, but the flat
depositional bankfull feature appears to be poorly
defined and the associated floodplain appears to be
very narrow. Based on this depiction, it might be
difficult to convince someone that the active
floodplain associated with bankfull discharge is easy
to identify. It also looks like a lot of additional water
will be needed before the flow leaves the channel
and inundates adjacent lands.
In contrast, the 1:1 cross-section plot (Figure 4)
conveys an alternative view of the river channel.
From the true-scale plot, the South Fork Cache la
Poudre River appears to be fairly wide and shallow,
rather than deep and narrow. The river appears to
be easily wadeable with gradual changes in depth.
The depositional flat of the floodplain associated
with bankfull stage is much more distinct and spans
a wide band along the edge of the river. Small
increases in stage above bankfull will quickly
inundate adjacent riparian ecosystems. In essence,
this plot shows the dimensions of the river as one
would perceive it standing along its banks.
It is almost always wise to view cross-sections at a
1:1 ratio to gain an appreciation for how they
actually look spatially. Most spreadsheet graphics
allow users to compress the vertical scale of the plot
and to manually resize the vertical scale. Some
software programs, such as WinXSPRO, allow users
to change the plotting ratio so that the cross-section
can be viewed at 1:1 and other scales of vertical
exaggeration. When rivers are wide, it may not be
possible to show cross-sections at a 1:1 ratio. In
these cases, consider using the lowest ratio feasible
and note on the figure or in the caption the amount
of vertical exaggeration used.
Cameras are another effective way to show what
channel cross-sections look like. Panoramic view
cameras, including inexpensive disposable versions,
are now widely available and are ideal for capturing
the width of the entire cross-section. Figure 5 is an
example of a panoramic view of the South Fork
Cache la Poudre cross-section.
By using these simple techniques, we can better
communicate the true nature of rivers to our various
audiences.
John Potyondy and Larry Schmidt are
hydrologist and Program Manager, respectively,
with the Rocky Mountain Research Station,
Stream Systems Technology Center;
(970) 498-1731; jpotyondy/rmrs@fs.fed.us;
lschmidt/rmrs@fs.fed.us
STREAM
NOTES
STREAM SYSTEMS TECHNOLOGY CENTER
USDA Forest Service
Rocky Mountain Research Station
240 West Prospect Road
Fort Collins, CO 80526-2098
IN THIS ISSUE
• Improving Future
Fluvial
Classification
Systems
• Why Do We
Exaggerate
Stream Channel
Cross-Section
Plots? The Case
for True Scale
Plotting
Rocky Mountain Research Station General Technical Report RM­
245, Stream Channel Reference Sites: An Illustrated Guide to
Field Technique by Cheryl C. Harrelson, C.L. Rawlins, and John
P. Potyondy is now available on the Internet. The publication is
a practical guide to establishing permanent reference sites for
gathering data about the physical characteristics of streams and
rivers. It describes procedures for selecting and mapping a site,
measuring channel cross-sections, surveying a longitudinal
profile, identifying bankfull stage, and measuring streamflow
and bed material. The guide, in pdf format, can be dowloaded
from the STREAM Web site (www.stream.fs.fed.us) FTP
download area. Printed copies of the 61 page report can be
obtained from the Rocky Mountain Research Station.
The United States Department of Agriculture (USDA) prohibits discrimination in its programs on the basis of race, color, national origin, sex,
religion, age, disability, political beliefs, and marital or familial status. (Not all prohibited bases apply to all programs.) Persons with
disabilities who require alternative means for communication of program information (braille, large print, audiotape etc.) should contact
USDA’s Target Center at 202-720-2600 (voice or TDD).
To file a complaint, write the Secretary of Agriculture, U.S. Department of Agriculture, Washington, DC 20250, or call 1-800-245-6340 (voice)
or 202-720-1127 (TDD). USDA is an equal employment opportunity employer.